Trim borders

Trim finds the uniform border around an image — the white margin on a screenshot, the empty edge on a scan — and crops it away. The border colour is taken from whichever of the four corners repeats most often, then each edge walks inward until it meets a row or column holding a pixel that differs by more than your tolerance, which runs from 0 (exact) to 64. Each file is measured once when you add it, so dragging the tolerance re-detects the edges without reading the pixels again.

How it works

  1. Drop your images Files are measured as they arrive. The readout in the rail fills in with the detected content box, a diagram of it, and the pixels cut from each edge.
  2. Set the tolerance Zero means an exact colour match. Raise it for scans and photographs, where the border is nearly but not quite one colour. It starts at 8.
  3. Decide how transparency counts "Treat transparent as empty" is on by default, so fully transparent pixels are border whatever colour sits under them. Turn it off to judge the border by colour alone.
  4. Add padding if you want it Positive padding leaves pixels around the content; negative padding cuts into it, and the page tells you by how much.
  5. Process and download Press process, then download. One file saves directly; a batch arrives as a zip.

Questions

What happens if an image has no uniform border?
Nothing is cut, and the page says so by name — "has no uniform border at a tolerance of 8" — and suggests raising the tolerance so near-matches count as border. The file still goes through the pipeline, so a format change applied to the batch is not silently skipped for it.
Is there a size limit on what can be trimmed?
Yes. Detecting a border needs the whole image in memory at once, so anything over 32 megapixels is not measured and passes through unchanged, with a message giving its megapixel count. An image that is a single flat colour keeps one pixel rather than becoming an empty file.
How does it work out which colour the border is?
It reads all four corners and takes whichever colour repeats most, with a four-way tie going to the top-left. A single stray pixel in one corner — a rounded window shadow, a scanner artefact — then costs a pixel of crop instead of breaking the detection entirely.

Every tool